An AWS outage is when part of Amazon Web Services stops working properly. So many apps run on AWS that one fault can take down sites that seem to have nothing to do with each other.
Regions and Availability Zones
AWS runs its cloud in Regions: separate geographic areas, such as us-east-1 in Northern Virginia. Each Region is split into Availability Zones, each one or more data centres with their own power and networking, kept apart so that a power cut in one doesn't take out the others.
That layout decides how far a failure reaches. If one Availability Zone goes down, only apps running in that zone alone are hit. If a service breaks across a whole Region, everything there that relies on it is hit, however many zones it uses.
The October 2025 us-east-1 outage
Late on 19 October 2025, Pacific time, AWS's us-east-1 Region started failing, and the disruption lasted until the afternoon of 20 October. AWS's post-event summary traced it to DynamoDB, its managed database service.
Two of the automated workers that keep DynamoDB's DNS records up to date got in each other's way. One applied an old plan, the other cleaned that plan up, and the regional endpoint, dynamodb.us-east-1.amazonaws.com, was left with an empty DNS record. Apps and AWS's own services could no longer look up the address to connect to. That is a race condition: two processes acting on the same thing, with the result depending on which finished first.
DynamoDB recovered in about three hours. The outage lasted much longer because other AWS services depend on it: new EC2 instances failed to launch for hours, Network Load Balancer health checks failed and took working capacity out of service, and Lambda saw errors and delays.
Why one fault spreads
Cloud services are built on other cloud services. When a shared building block such as a database or DNS fails, everything above it fails too, even when its own code is fine. DNS shows why: if a name can't be turned into an address, a healthy server is unreachable. Any distributed system has to plan for one of its parts failing like this.
What teams can do about it
- Run across several Availability Zones, so losing one data centre isn't an outage. That would not have helped in October 2025, because the fault covered the whole Region.
- For the systems that matter most, plan for a second Region and practise failing over to it. It costs more: two sets of infrastructure, and data to keep in sync.
- Fail gracefully: show cached data or a clear message rather than a blank page.
- Know your dependencies, including the services your provider relies on.
Key takeaways
- AWS splits its cloud into Regions, and each Region into isolated Availability Zones.
- In October 2025, an empty DNS record took DynamoDB down in
us-east-1, and the failure spread to EC2, load balancers and Lambda. - Several Availability Zones protect you from one data centre failing; only a second Region protects you from a whole Region failing.
Wondering how this was made? How we make these